Continuing Tracks
AI Safety Continuing tracks are open to those who have completed an AI Safety intro fellowship, or have otherwise demonstrated an understanding of core AI safety concepts. We offer both upskilling tracks and reading groups.
Upskilling tracks are intensive quarter-long programs designed to help you develop practical, hands-on skills in cutting-edge AI technologies. These tracks focus on building technical expertise in specialized areas that are crucial for AI safety research, providing you with the concrete tools and implementation experience needed to contribute to the field. Currently we offer an 8-week CUDA Track, in which participants will master parallel computing with CUDA by engaging in technical readings and completing fun implementation assignments outside of hour-long weekly meetings.
Reading groups are quarter-long programs designed to delve further into relevant AI safety concepts and research with other engaged participants. We offer two reading groups. The first reading group is built around the bestselling book, If Anyone Builds It, Everyone Dies (IABIED): we will interrogate the claims made in the book and discuss the implications of the book’s recommendations for how society manage AI risk. The second reading group is our technical paper reading group: each week participants will read a cutting-edge paper in ML or AI safety and discuss to build a more comprehensive understanding of technical AI safety.
Curricula: